[MUD-Dev] Reputation systems: a possible path for investigation

[J C Lawrence]

On Mon, 11 Aug 2003 01:09:42 -0400 
david l smith <david.l.smith at mail-x-change.com> wrote:
> On Sunday, August 10, 2003, J C Lawrence said

>> Reputation systems are usually thought of as interesting ways for the
>> game system to define and provide predetermined and service-qualified
>> hinting and decision making keys to players about other players (PK,
>> trade, trust etc).

> Actually, as a developer, I was originally more interested in a
> reputation system that would let me add a noosphere {The article
> mentioned "http://www.firstmonday.org/issues/issue7_7/masum/" by
> Hassan Masum uses this notion, and it fits much better than my
> previous thinking. 

I don't think we're saying anything different.  Quoting myself:

  > Arbitrary graining and subjective contextual per-player
  > interpretation...

and:

  > A possible approach instead is to treat the players as a homogeneous
  > mesh and allow them to define the nouns and verbs in the reputation
  > system, and possibly some of the high level grammars.

That pretty much defines a system where the users, players, define the
terms that have reputations attached to them, the definitions of those
terms, their semantics, value sets and ranges, the grammar with which
they are constructed etc.  

How would that differ from the sort of noospwhere that you refer to?

> In a simulated environment, a part of the state of that simulation is
> the body of experience in the system: things that "everyone knows".

I think part of the problem is a confusion over what a reputation system
is, and what the payloads are that are carried by a reputation system.
I take a broad view, deriving from the definition of reputation:

  Reputation (n): The character attributed to a person, thing, or
  action; repute.

As such almost any conclusion that can be derived from a game, its
contents, or its players, or any player "think" about anything in a game
that can be put into a declarative statement is a valid payload for a
reputation "message".

More simply I view a reputation system as an undefined thing (in the
same sense that the TCP/IP "cloud" is an undefined "thing") that can
deliver hinting structures to players for their decision making
processes, no matter what, or how useful, the hints provided are.

There are several somewhat overlapping data sets which can be referred
to by the term, "reputation system":

  1) An auditable recorded trail as to a thing's past behaviour as
  evaluated by defined metrics and reported in a standard fashion.  eg
  UO's and M59's flagging of PKers.

  2) An inter-player communication of perceived behaviour patterns,
  often partially supported, reported in a contextual and subjective
  fashion.  eg Almost all player interaction or knowledge-transfer as
  regards a game, including guild/group agglomerations and other proxy
  forms.

  3) Gossip.  "Did you hear that...?"

  4) Rumours.  "I heard from someone who heard that..."

The last two are similar but bear distinct differences in that rumours
are explicitly third hand, can be machine simulated (gossip can't), and
suffer fairly well known and understood decay and mutation curves.
Gossip is always first hand (even when reported as third person
transcription), can't be machine simulated, travels well known
distribution paths that are extremely hard to simulate, and has equally
complex and discontinuous decay and mutation curves.

In the general, ultra-generic case I think of them as all flavours of
reputation systems as they all represent, in their transit, what a
player thinks of something -- what the reputation of the subject is with
the speaker.

  Bubba is a PK'er.  

  The rotting rope bridge is safe to cross.  

  Guild Boffo are idiots.

  Boffo PKed Bernie, Bubba, and Bruce.

  Bernie has been PKed five times.

  A good place to dig for mushrooms is around Fortress Fract.

  The orcs by the western caverns are push overs.

  Carrying a glass cross seems to make your more resistant to damage in
  melees.

  You get better quality iron ore if you spend a quarter of your time
  digging with a wooden shovel and the rest of the time with a pick.

  Stand behind the boulder when fighting the troll by the bridge and
  you'll get hit less.  

  Boffo has the hots for Bernie.

  Bubba's gonna kick Bruce's arse.

  If you walk backwards you can get closer to a dragon before you get
  hit.

  There's a lot of gold in Castle Krak.

  The elven sceptre does something weird in Castle Krak.

  Watch out for the balrog.

All of these statements are hinting items which can cause and affect the
decisions players make.  They are sample payloads which can be
communicated by a reputation system.

> That this includes much more than "faction a reaction bonus towards
> faction b" isn't well conveyed by most systems that I've seen). While
> the per-faction reputation tracking systems that today boil down to
> the one-dimensional relationship of "reaction bonus" are better than
> nothing, a more free-form (or rather, higher dimensionality of
> comparison without much more code) approach could make for some
> interesting emergent characteristics from the state of the noosphere.

Quite.

Its another thing to just hand the whole potato over to players and let
them determine what is going to be reported, how, on what scales, using
what nouns, etc.

Brian (as he's already posted) is somewhat concerned about the impact on
knowledge brokers.  In brief discussion of this background with Jon
Leonard  on DevMUD (along with brain as happens), Jon raised the issues
of trust, abuse, propaganda, and collateral damage.  

  What happens when the (majority of the?) data on your isn't
  verifiable, or, simply, can't have its authorship perfectly traced?

  What do the players do and how do they handle such a reputation
  system?

  What intersections and impacts do reputation systems have on the
  timeliness and locality of data?  

Can a player or band of players inject sufficient distorted and
self-serving data into the reputation system (by whatever means) that it
effectively becomes a propaganda machine and an instituter of the Big
Lie?  Will players use and even trust such a necessarily untrustworthy
reputation system, and if not, will they create information economies to
derive value using it as one resource among several?  Can your game
design survive a subjectively evaluated lack of secrets?  Can your
player cultures?  Can you be sued for libel or defamation of character
for lies or maligning statements reported on a game's reputation system?

>   -- All non-player characters in the simulation form opinions...
>   -- When non-player characters interact with each other...

Bingo.

>   -- With some framework to cross correlate context data...

While I like the idea of the system providing sample correlations as a
seed culture, is it really necessary in the long run?

>   -- If you wanted to finish the circle, tie it into your player
> character's experience, depending on the era, allow them to buy a
> newspaper, or "chat up the barkeep" to get a little slice of the hot
> gossip in the system nooosphere. Might be nice to know that the local
> keep thinks that the guy you're doing business with is shady, might be
> worth your silver to learn that he's rumored to have killed three good
> men. Then again, maybe you're fine going off in his party with him,
> never can tell really...

  > go north
  You approach a chasm which is crossed by a rotting rope bridge.
  > examine bridge
  It is made of rope and looks in very poor shape.  The rope is visibly
  rotten in places and there are several large holes in which look like
  they were made by something falling through the bridge.
  > consult system about bridge
  Bubba says, "The bridge is safe."  Score: 37.5
  Boffo says, "Beware of the third step, its a doozie!"  Score: 18.3
  Bernie says, "Make sure you have the glowing toadstool." Score: -3.7
  Bruce says, "Get thee across the bridge, varlet!" Score: -7.8
  ...etc

--
J C Lawrence                
---------(*)                Satan, oscillate my metallic sonatas. 
claw at kanga.nu               He lived as a devil, eh?		  
http://www.kanga.nu/~claw/  Evil is a name of a foeman, as I live.

_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev